Pairwise Interactions between Neuronal a 7 Acetylcholine Receptors and a -Conotoxin ImI*

The present work uses a -conotoxin ImI (CTx ImI) to probe the neurotransmitter binding site of neuronal a 7 acetylcholine receptors. We identify key residues in a 7 that contribute to CTx ImI affinity, and use mutant cycles analysis to identify pairs of residues that stabilize the receptor-conotoxin complex. We first mutated key residues in the seven known loops of a 7 that converge at the subunit interface to form the ligand binding site. The mutant subunits were expressed in 293 HEK cells, and CTx ImI binding was measured by competition against the initial rate of 125 I- a -bungarotoxin binding. The results reveal a predominant contribution by Tyr-195 in a 7 , accompanied by smaller contributions by Thr- 77, Tyr-93, Asn-111, Gln-117, and Trp-149. Based upon our previous identification of bioactive residues in CTx ImI, we measured binding of receptor and toxin mutations and analyzed the results using thermodynamic mutant cycles. The results reveal a single dominant interaction between Arg-7 of CTx ImI and Tyr-195 of a 7 that anchors the toxin to the binding site. We also find multiple weak interactions between Asp-5 of CTx ImI and Trp-149, Tyr-151, and Gly-153 of a 7 , and between Trp-10 of CTx ImI and Thr-77 and Asn-111 of a 7 . The overall results establish the orientation of CTx ImI as it bridges the subunit interface and demonstrate close approach of residues on opposing faces of the a 7 binding site. By

binding sites of muscle AChRs, as well as differentiate muscle from neuronal AChRs, ␣-conotoxins have become valuable probes of agonist binding sites of the nicotinic AChR superfamily (4 -8). As an example of ␣-conotoxin selectivity for AChR subtypes, the ␣-conotoxins MI and GI bind with high affinity to muscle but not neuronal AChRs, whereas ␣-conotoxins ImI and MII bind with high affinity to certain neuronal but not muscle AChRs (9 -11). All ␣-conotoxins contain N-and C-terminal loops, two disulfide bridges, and proline in the N-terminal loop, but the various ␣-conotoxins differ by the composition and number of residues in each loop (Fig. 1). Thus, structural differences among ␣-conotoxins and AChR binding sites determine the specificity of receptor-conotoxin interactions.
Mutagenesis and site-directed labeling studies establish that the ligand binding sites of the AChR are formed at interfaces between subunits. Whereas the two binding sites of muscle AChR are formed at interfaces between ␣ 1 and either ␦ or ␥ subunits, binding sites of the ␣ 7 neuronal AChR are formed at interfaces between identical ␣ 7 subunits. Residues of the ␣ face of the binding site, termed the (ϩ) face, cluster in three well separated regions of the primary sequence, termed loops A, B, and C (Fig. 2). Using the numbering system for human ␣ 7 receptors, key residues in these loops include Tyr-93 in loop A, Trp-149 in loop B, and Tyr-188 and Tyr-195 in loop C (12,13). Similarly, residues of the non-␣ face of the binding site, termed the (Ϫ) face, cluster in four separate regions of the primary sequence, termed loops I through IV (Fig. 2). Key residues in these loops include Ser-34 in loop I, Trp-55 in loop II, Gln-117 in loop III, and Asp-164 in loop IV (12,13). The observation that these seven loops converge to form a localized binding site has led to a multiloop model of the major extracellular domain of the AChR (12).
Previous work used the competitive antagonist dimethyl-dtubocurarine (DMT) to probe the ligand binding site of the muscle AChR (14). Those studies showed that residues on opposing faces of the subunits, Tyr-198 in the ␣ subunit and Tyr-117 in the ␥ subunit, approach to within 11 Å to provide aromatic donors that stabilize the two positively charged nitrogens of DMT. ␣-Conotoxins permit analogous studies of binding site structure, but unlike DMT or small organic ligands, they can be readily modified using peptide synthesis. Owing to their larger size, asymmetric structure, and probable multiple points of contact, ␣-conotoxins should lead to further definition of the architecture of the AChR binding site.
We previously identified residue differences between ␣ 7 and muscle AChRs that confer neuronal specificity for CTx ImI (15). In addition, we identified residues in CTx ImI that determine its affinity for ␣ 7 receptors (16). The present work identifies pairs of residues that stabilize the complex formed between ␣ 7 receptors and CTx ImI. We find that the receptorconotoxin complex is stabilized by a dominant aromaticquaternary interaction, supplemented by multiple weaker interactions. The results establish the orientation of CTx ImI as it bridges the subunit interface.

EXPERIMENTAL PROCEDURES
Materials-125 I-Labeled ␣-bungarotoxin (␣-Bgt) was purchased from NEN Life Science Products, d-tubocurarine chloride from ICN Pharmaceuticals, 293 human embryonic kidney cell line (293 HEK) from the American Type Culture Collection, and unlabeled ␣-Bgt from Sigma Chemicals. Human ␣ 7 and rat 5HT-3 subunit cDNAs were generously provided by Drs. John Lindstrom and William Green.
Synthesis and Purification of Conotoxin ImI-Wild-type and mutants of ␣-conotoxin ImI were synthesized by the Mayo Protein Core Facility using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry on an Applied Biosystems 431A peptide synthesizer. During synthesis, cysteine (S-triphenylmethyl)-protecting groups were incorporated at cysteines 3 and 12, and acetamidomethyl-protecting groups were incorporated at cysteines 2 and 8. The linear peptide was purified by reversed phase HPLC using a Vydac C18 preparative column with trifluoroacetic acid/acetonitrile buffers. The two intramolecular disulfide bridges were formed as follows; the S-triphenylmethyl protecting groups attached to cysteines 3 and 12 were removed during trifluoroacetic acid cleavage of the linear peptide from the support resin. The peptide was then oxidized by molecular oxygen to form the 3-12 disulfide bond by stirring in 50 mM ammonium bicarbonate buffer, pH 8.5, at 25°C for 24 h. The peptide was lyophilized prior to formation of the second disulfide bond. The acetamidomethyl-protecting groups on cysteine 2 and 8 were removed oxidatively by iodine as described (17), except the peptide/iodine reaction was allowed to progress for 16 h prior to carbon tetrachloride extraction. The pure product was separated from residual iodine by HPLC, and verified by mass spectrometry. Formation of disulfide bonds was confirmed as described previously (16). The CTx ImI mutants are named as follows; the first letter and number refer to the wild-type residue and position, and the following letter is the substituted residue at that position.
Mutagenesis and Expression in HEK Cells-AChR subunit cDNAs were subcloned into the cytomegalovirus-based expression vector pRBG4 (18). Mutant cDNAs were constructed by bridging naturally occurring or mutagenically installed restriction sites with doublestranded oligonucleotides or by the Quick Change TM site-directed mutagenesis kit (Stratagene). All constructs were confirmed by dideoxy sequencing. To increase expression, the extracellular domains of ␣ 7 and all mutations were joined to the rat 5HT-3 sequence at the start of the M1 transmembrane domain (15). The ␣ 7 point mutants are named as follows; the first letter and number refer to the wild-type residue and position, and the following letter is the substituted residue at that position. HEK cells were transfected with either wild-type or mutant cDNAs using calcium phosphate precipitation as described (18). For ligand binding measurements, cells 2 days after transfection were harvested by gentle agitation in phosphate-buffered saline (PBS) containing 5 mM EDTA.
Ligand Binding Measurements-Ligand binding to intact cells was measured by competition against the initial rate of 125 I-␣-Bgt binding (18). The cells were briefly centrifuged, resuspended in potassium Ringer's solution, and divided into aliquots for ligand binding. Potassium Ringer's solution contains: 140 mM KCl, 5.4 mM NaCl, 1.8 mM CaCl 2 , 1.7 mM MgCl 2 , 25 mM HEPES, and 30 mg/liter bovine serum albumin, adjusted to a pH of 7.4 with NaOH. Specified concentrations of wildtype or mutant CTx ImI were added 30 min prior to addition of 3.75 nM 125 I-␣-Bgt, which was allowed to bind for 10 min to occupy approximately half of the surface receptors. Binding was terminated by addition of 2 ml of potassium Ringer's solution containing 600 M d-tubocurarine chloride. All experiments were performed at 24 Ϯ 2°C. Cells were harvested by filtration through Whatman GF-B filters using a Brandel Cell Harvester and washed three times with 3 ml of potassium Ringer's solution. Prior to use, filters were soaked in potassium Ringer's solution containing 4% skim milk. Nonspecific binding was determined in the presence of 10 nM ␣-Bgt and was typically 1% of the total number of binding sites. The total number of binding sites was determined by incubation with toxin for 120 min. The initial rate of toxin binding was calculated as described (19) to yield the fractional occupancy of competing ligand. Binding measurements were analyzed according to the monophasic Hill equation (Equation 1).
Y is fractional occupancy of the competing ligand, K app is the apparent dissociation constant, and n H is the Hill coefficient. Fitted parameters and standard errors were obtained using UltraFit (BIOSOFT). For multiple experiments, means of the individual fitted parameters and standard deviations are presented (Tables I and III).

Mutagenesis of the (ϩ)
Face of the ␣ 7 Binding Site-Before we could identify pairs of residues that stabilize the ␣ 7 -CTx ImI complex, we first needed to identify residues in ␣ 7 required for CTx ImI binding. Beginning with the (ϩ) face of the ␣ 7 binding site, we mutated key aromatic residues in loops A, B, and C, as well as local flanking residues in each loop (Fig. 2). To determine functional consequences of the mutations, each cDNA was transfected into 293 HEK cells and binding of CTx ImI was measured by competition against the initial rate of 125 I-␣-Bgt binding. To achieve high expression in 293 HEK cells, all mutations were generated in the ␣ 7 /5HT-3 chimera, which consists of ␣ 7 sequence from the N terminus to the M1 transmembrane domain followed by 5HT-3 sequence to the C terminus (15).
We find that Tyr-195 in loop C provides the dominant source of stabilization of CTx ImI, as shown by the 320-fold loss in affinity produced by Y195T ( Fig. 3 and Table I). Similarly, mutation of Tyr-195 to arginine decreases affinity 2500-fold, probably through electrostatic repulsion against one of the three cationic moieties in CTx ImI. On the other hand, mutation of Tyr-195 to phenylalanine does not affect affinity, indicating that an aromatic side chain is required. In addition to Tyr-195, we find that residues in loops A, B, and C contribute to CTx ImI binding, showing 10-to 30-fold changes in affinity for the various mutations; these include Y93T, W149T, Y151T, G153S, R186V, and D197N ( Fig. 3 and Table I). Thus, on the (ϩ) face of the binding site, Tyr-195 makes the dominant contribution to CTx ImI affinity, while additional residues in loops A through C make smaller contributions.
Mutagenesis of the (Ϫ) Face of the ␣ 7 Binding Site-We next examined key residues in loops I through IV of the (Ϫ) face of the binding site (Fig. 2). In contrast to the (ϩ) face, residues of the (Ϫ) face contribute only weakly to CTx ImI binding (Fig. 4). The greatest contributions are found in loops II and III, including Thr-77 in loop II and Asn-111, Gln-117, and Pro-120 in loop III. We previously showed that Thr-77 contributes to CTx ImI specificity for ␣ 7 neuronal over muscle receptors (15). The mutation N111S produces an unusual increase rather than decrease in affinity, presumably due to removal of glycosylation. In support of a glycosylation site, mutating the downstream consensus site with S113A has the same effect as N111S. Removing carbohydrate from Asn-111 may increase affinity for CTx ImI by allowing deeper penetration into the binding site. Gln-117 occupies a position equivalent to Tyr-117 in the muscle receptor ␥ subunit, which stabilizes one of two quaternary nitrogens in DMT (14,18). Finally, Pro-120 is the first of a pair of vicinal prolines, both highly conserved, that likely stabilize the conformation of loop III. Thus the (Ϫ) face of the binding site stabilizes CTx ImI through relatively weak contributions by residues in loops II and III.
Effects of the Mutations on ␣-Bgt Binding-Because CTx ImI binding is measured by competition against 125 I-␣-Bgt binding, we determined whether the mutations in ␣ 7 affect binding of the reporter ligand itself. We therefore compared rates of 125 I-␣-Bgt association (k T ) for wild-type and mutant ␣ 7 receptors (Table II). Association rates for 125 I-␣-Bgt are similar for wildtype and mutant ␣ 7 receptors, indicating that residues contributing to CTx ImI binding do not affect entry of ␣-Bgt to the site.
To summarize, we have examined key residues in the seven loops in ␣ 7 known to contribute to the ligand binding site. We find that Tyr-195 is the major contributor to CTx ImI affinity, and that multiple residues on both the (ϩ) and (Ϫ) faces of the binding site are minor contributors.
Pairs of Interacting Residues That Stabilize the Receptor-Conotoxin Complex-We previously identified key residues in each of the two loops of CTx ImI that contribute to its bioactivity (Fig. 1). In the N-terminal loop, each of three consecutive residues, Asp-5, Pro-6, and Arg-7 is equally important for bioactivity, suggesting that they contribute as an interdependent triad (16). Systematic mutagenesis of each residue indicates that this bioactive triad requires conformational restriction by Pro-6 and proper chain length of the oppositely charged side chains. In the C-terminal loop, Trp-10 contributes to bioactivity through its aromatic ring. Thus, knowing key residues in both CTx ImI and ␣ 7 , we designed experiments to identify pairs of residues that stabilize the receptor-conotoxin complex.
To identify pairs of interacting residues, we focused on residues shown by mutagenesis to provide significant stabilization of the receptor-conotoxin complex. Results for the most strongly interacting pair of residues are shown in Fig. 5A, which displays binding curves for the four possible combinations of wildtype and mutant for the receptor-conotoxin pair Y195T/R7Q (Fig. 5A). The receptor mutation Y195T reduces affinity for wild-type CTx ImI by 320-fold, while the CTx ImI mutation R7Q reduces affinity for wild-type ␣ 7 by 380-fold. However when examined together, the two mutations show no further decrease in affinity, contrary to the 10,000-fold decrease expected if their contributions were additive.
On the other hand, results demonstrating non-interacting residues are displayed in Fig. 5B for the receptor-conotoxin pairY93T/W10T. The receptor mutation Y93T decreases affinity for wild-type CTx ImI by 40-fold, while the CTx ImI muta-

FIG. 3. Mutagenesis of residues in loops A through C of the (؉)
face of the ␣ 7 receptor binding site. 293 HEK cells were transfected with wild-type and mutant ␣ 7 /5HT-3 subunit cDNAs, and binding of CTx ImI to the resulting surface receptors was determined as described under "Experimental Procedures." The resulting dissociation constants are expressed relative to that of wild-type ␣ 7 as the log ratio. The vertical line indicates the ratio for wild-type ␣ 7 and CTx ImI, with the error bars indicating Ϯ S.D. Means and standard errors of the dissociation constants are given in Table I.  Table I. NE indicates no expression. tion W10T decreases affinity for wild-type ␣ 7 by 30-fold. The two mutations together decrease affinity by 700-fold, which is purely additive, indicating that Tyr-93 and Trp-10 do not interact. Thus analyses of pairwise mutations in ␣ 7 and CTx ImI readily distinguish interacting from non-interacting residues.
Thermodynamic Mutant Cycles Analysis-Thermodynamic mutant cycles analysis has been widely used to identify interactions between pairs of residues in proteins, as well as to estimate the free energy of the interactions (1-3, 20, 21). Generating a mutant cycle requires measurements of dissociation constants (K d ) for the four possible combinations of wild-type (W) and mutant (M) receptors (r) and toxins (t): WrWt, MrWt, WrMt and MrMt. The resulting set of dissociation constants are then used to calculate a coupling coefficient ⍀.
When ⍀ is unity, the pair of residues does not interact, whereas, when ⍀ deviates significantly from unity, the pair of residues interacts. The free energy of the interaction is given by Equation 3.
Mutant Cycles Analysis of Residues on the (ϩ) Face of ␣ 7 -We

I-␣-Bgt association with wild-type and mutant ␣ 7 /5HT-3 receptors
Data are rate constants for 125 I-␣-Bgt association (k T ) for the indicated ␣ 7 point mutants, expressed relative to wild-type ␣ 7 /5HT-3 receptors. Rates are calculated from the bimolecular rate equation: where T o and R o are the initial concentrations of toxin and receptors, respectively, and RT is the amount of toxin-receptor complex formed during the interval t. For wild-type ␣ 7 /5HT-3 receptors k T ϭ 5.83 ϫ 10 5 M Ϫ1 s Ϫ1 . next examined all residues on the (ϩ) face of the ␣ 7 binding site that significantly stabilize CTx ImI. Each receptor mutation was tested against a single conotoxin mutation, and the process was repeated for each of the four key residues in CTx ImI (Fig.  1). Among all possible combinations of receptor and conotoxin mutations, the pair Y195T/R7Q shows the largest coupling coefficient of 325, corresponding to a free energy of interaction of 3.4 kcal/mol (Fig. 6, Table III). Both the side chain chemistry and the coupling free energy suggest that Tyr-195 and Arg-7 form a -cation interaction that provides the primary anchor of the receptor-conotoxin complex. Tyr-195 also couples significantly to the remaining two residues of the CTx ImI triad (Fig. 6, Table III). The pair Y195T/ P6G shows a coupling coefficient of 46, suggesting that release of the conformational restriction by proline allows greater mobility of the adjacent Arg-7. Finally, the pair Y195T/D5N shows a small but significant coupling coefficient of 10, suggesting that Asp-5 helps position Arg-7 through electrostatic attraction.
The coupling for the Arg-7/Tyr-195 and Pro-6/Tyr-195 pairs is specific because the remaining key residues on both the (ϩ) and (Ϫ) face show only weak or no coupling to either Arg-7 or Pro-6 ( Figs. 6 and 7). Asp-5, on the other hand, also couples to residues in loop B of the (ϩ) face, including Trp-149, Tyr-151, and Gly-153 (Fig. 6, Table III). The coupling between Asp-5 and FIG. 5. Binding of wild-type and mutant ␣ 7 and CTx ImI pairs. A, 293 HEK cells were transfected with wildtype ␣ 7 /5HT-3 or ␣ 7 Y195T subunit cDNAs, and binding of wild-type CTx ImI or R7Q mutant CTx ImI was determined as described under "Experimental Procedures." The curves through the data are fits to the Hill equation (Equation 1), with means and standard errors of the fitted parameters given in Table I and III. The drawing at the top is a schematic representation of CTx ImI with the bioactive residues in boldface text and arrow indicating the R7Q mutation. B, 293 HEK cells were transfected with wild-type ␣ 7 / 5HT-3 or ␣ 7 Y93T subunit cDNAs, and binding of wild-type CTx ImI or W10T mutant CTx ImI was determined as described under "Experimental Procedures." The curves through the data are fits to the Hill equation (Equation 1), with means and standard errors of the fitted parameters given in Table I and III. The drawing at the top is a schematic representation of CTx ImI with the bioactive residues in boldface text and arrow indicating the W10T mutation.  Table III.
Points of Contact between ␣ 7 Receptors and ␣-Conotoxin ImI residues in loop B appears specific because the remaining binding site residues show no coupling to Asp-5. Thus, residues in loops B and C of ␣ 7 are positioned to interact with the bioactive triad in CTx ImI, with Tyr-195 and Arg-7 serving as the major anchor for stabilizing the complex. The overall results establish the orientation of bound CTx ImI: the N-terminal loop of CTx ImI orients toward the (ϩ) face of the ␣ 7 binding site.
Mutant Cycles Analysis of Residues on the (Ϫ) Face of ␣ 7 -Contrary to the strong coupling between CTx ImI and the (ϩ) face of the ␣ 7 binding site, we find only weak coupling between CTx ImI and residues on the (Ϫ) face (Fig. 7). The weak coupling mirrors the relatively weak stabilization provided by residues on the (Ϫ) face (Fig. 4). The strongest coupling occurs between Trp-10 in the C-terminal loop of CTx ImI and Thr-77 and Asn-111 in loops II and III of ␣ 7 , respectively (Table III and Fig. 7). The coupling to Asn-111 results from removal of carbohydrate that likely improves steric fit of Trp-10 as it protrudes from the toxin scaffold. We previously identified Thr-77 as one of three determinants of CTx ImI specificity for ␣ 7 neuronal over muscle receptors (15). When all three ␣ 7 specificity determinants were mutated in a single receptor (W55R/S59Q/T77K), no additional coupling was detected to Trp-10 (Table III).
The coupling of Trp-10 to the (Ϫ) face of ␣ 7 is specific because neither Asp-5 nor Arg-7 in the N-terminal loop of CTx ImI couple to residues on the (Ϫ) face (Fig. 7). On the other hand, Pro-6 in the N-terminal loop couples to Thr-77 and Asn-111 on the (Ϫ) face of ␣ 7 (Fig. 6). Because Pro-6 and Trp-10 are well separated, their coupling to the same binding site residues suggests that the mutation P6G increases flexibility that spreads from the N-to the C-terminal loop of the toxin. These results, together with conservation of Pro-6 in all ␣-conotoxins (Fig. 1), suggest that Pro-6 maintains rigidity of the toxin along with the two disulfide bridges. Thus, within the (Ϫ) face of the ␣ 7 binding site, loops II and III appear positioned to interact with Trp-10 in the C-terminal loop of CTx ImI. The overall results establish that CTx ImI bridges the subunit interface formed by pairs of ␣ 7 subunits; the N-terminal loop of the toxin orients toward the (ϩ) face of ␣ 7 , while the C-terminal loop orients toward the (Ϫ) face. DISCUSSION By identifying pairs of interacting residues that stabilize the ␣ 7 -CTx ImI complex, the present work clarifies our picture of AChR binding site structure and provides insight into the nature of AChR-ligand interactions. We find that CTx ImI anchors to the ␣ 7 binding site primarily through interaction between Arg-7 in its N-terminal loop and Tyr-195 on the (ϩ) face of the binding site. This primary interaction is accompanied by multiple weaker interactions between Asp-5 in the N-terminal loop of the toxin and additional residues on the (ϩ) face of ␣ 7 , and between Trp-10 in the C-terminal loop of the toxin and loops II and III of the opposing (Ϫ) face of ␣ 7 . The residues identified show little or no coupling to the many theoretically possible alternative residues, indicating that the coupling is specific and the pairs of residues are points of contact in the ␣ 7 -CTx ImI complex.
Mutagenesis studies have provided considerable insight into key residues in the AChR that contribute to ligand binding (13). However, it is often difficult to establish that a key residue interacts directly with a ligand because interactions with other residues or propagated effects are difficult to rule out. By systematically mutating both the ligand and the receptor and applying mutant cycles analysis, the likelihood of establishing a direct interaction is very high (20). Although our observation of a large coupling coefficient between Tyr-195 and Arg-7 does not guarantee that the two residues interact directly, the specificity of the coupling strongly indi-cates that they do. Alternatively, if Tyr-195 interacted with a second residue in ␣ 7 that in turn contacted Arg-7, mutating that second residue should both weaken CTx ImI binding and show strong coupling to Arg-7. Our systematic mutagenesis shows no coupling of Arg-7 to other key residues of either the (ϩ) or the (Ϫ) face of the binding site, even though CTx ImI binding is very sensitive to mutation of any of these key residues. Furthermore, Tyr-195 couples to all three members of the interdependent triad in the N-terminal loop of CTx ImI, but not to Trp-10 in the C-terminal loop. Thus, our results reveal points of direct interaction between CTx ImI and the ␣ 7 binding site.
We find additional weak interactions between CTx ImI and the (ϩ) face of the binding site. A stretch of three residues in loop B of the (ϩ) face interact with Asp-5 in the N-terminal loop of CTx ImI. As the three residues, Trp-149, Tyr-151, and Gly-153, occupy a local stretch of sequence and show similar coupling free energies of 1.4 kcal/mol, the interaction with Asp-5 likely involves either the ␣ 7 peptide backbone in this region or an interdependent structure formed by all three side chains. These results again support orientation of the N-terminal loop of CTx ImI toward the (ϩ) face of the binding site.
On the other hand, we find that Trp-10 in the C-terminal loop of CTx ImI couples to residues in loops II and III at the (Ϫ) face of ␣ 7 . In loop II, Thr-77 contributes to neuronal specificity of CTx ImI, while in loop III, Asn-111 forms a glycosylation site. The coupling of Trp-10 to both of these residues suggests that loops II and III of the (Ϫ) face come into close proximity to each other. The coupling of Trp-10 to Thr-77 and Asn-111 is specific because Trp-10 does not couple to other key residues at either the (ϩ) or (Ϫ) faces of ␣ 7 . Unexpectedly, Pro-6 in the N-terminal loop of the toxin also couples to Thr-77 and Asn-111. However, this coupling likely arises from release of conformational constraint by mutation of Pro-6 that spreads to Trp-10 in the C-terminal loop of the toxin. The total lack of coupling of either Asp-5 or Arg-7 to residues on the (Ϫ) face of the binding site further demonstrates specificity of the coupling between Trp-10 of CTx ImI and the (Ϫ) face of ␣ 7 .
Studies of the muscle AChR have demonstrated that its two ligand binding sites are generated by pairs of ␣ 1 and non-␣ 1 subunits (13). Analogously, ligand binding sites of the ␣ 7 homomer are generated by pairs of identical ␣ 7 subunits. Although chemical cross linking and mutagenesis studies suggest that both ␣ 1 and non-␣ 1 subunits interact with bound ligand (14,25), evidence that residues on opposing subunits interact with a single bound ligand has been scarce. By showing that Arg-7 of CTx ImI interacts with the (ϩ) face of the binding site, while Trp-10 interacts with the (Ϫ) face of the site, our studies establish the subunit interface concept at the level of bound ligand.
Studies of chimeric subunits have demonstrated the modular exchangeability of segments of the AChR subunits, and have lead to a multi loop model of the folding pattern of the major extracellular domain of the subunit (22,23). In particular, three loops (A-C) converge to the traditional ␣ or (ϩ) face, while four loops (I-IV) converge to the traditional non-␣ or (Ϫ) face (Fig. 2). The modular exchangeability extends to individual residues, suggesting that even for the evolutionarily distant subunits, ␣ 1 and ␣ 7 , equivalent positions of the primary sequence occupy equivalent positions in three-dimensional space (15). Our findings further support this basic scaffold hypothesis by showing that key binding site residues identified in muscle AChRs are also present at the binding site in ␣ 7 .
Homology modeling of muscle AChR subunits and bacterial metal-binding proteins has provided a working model of the atomic structure of the ligand binding site (12). It should be possible to model the ␣ 7 binding site by using the muscle AChR model as a template, together with constraints imposed by the CTx ImI scaffold and its points of contact with ␣ 7 . The distance between the bioactive groups Arg-7 and Trp-10 of CTx ImI defines the minimum distance between residues on opposing faces of the ␣ 7 binding site.
Members of the ␣-conotoxin family show a striking range of affinities for different subtypes of AChR, presumably owing to differences in primary sequences of the AChR subunits, as well as differences in the number and types of residues in the various ␣-conotoxins. In particular, ␣-conotoxins MI and GI have negligible affinity for ␣ 7 receptors (16), but they bind with nanomolar affinity to the ␣-␦ site of muscle receptors (4,26). Mutagenesis studies of the ␣-␦ interface revealed that single residues in four loops contribute to CTx MI binding: Tyr-198 in loop C of the ␣ 1 subunit (26), Ser-36 in loop I, Tyr-113 in loop III, and Ile-178 in loop IV of the ␦ subunit (4). Two of these residues, Tyr-198 and Tyr-113, are equivalent to the two primary points of contact between ␣ 7 and CTx ImI: Tyr-195 and Asn-111. Analogous mutant cycles studies between the muscle AChR and CTx MI are required to answer such questions as whether CTx MI orients similarly to CTx ImI when bound to its site, or whether any of the three positively charged nitrogens of CTx MI interact with Tyr-198 of the ␣ 1 subunit.
Binding of CTx ImI to the ␣ 7 AChR shares features common to binding of other well studied competitive antagonists. We find that CTx ImI anchors to the ␣ 7 binding site through a single dominant interaction between Arg-7 of CTx ImI and Tyr-195 of ␣ 7 . The preference for an aromatic side chain at position 195, together with destabilization by the mutation Y195R, point to an aromatic-quaternary interaction between Tyr-195 of ␣ 7 and Arg-7 in CTx ImI. The coupling free energy for the Y195T/R7Q pair is 3.4 kcal/mol, which is in the range observed for aromatic-quaternary interactions (24). Analogously, the equivalent residue in the muscle ␣ 1 subunit, Tyr-198, stabilizes one of two quaternary nitrogens in dimethyl-dtubocurarine (14). ACh, with its quaternary nitrogen also requires tyrosine at position 195 of ␣ 7 , but unlike CTx ImI and dimethyl-d-tubocurarine, shows reduced apparent affinity for the Y195F receptor (27,28). Moreover unlike CTx ImI, ACh binding depends much more on aromatic groups at other positions, including Tyr-188, Trp-149, and Tyr-93 in ␣ 7 (29). The question remains open whether these aromatic residues in the (ϩ) face of the subunit interface or residues in the opposing (Ϫ) face serve as the final docking site for ACh. The emerging picture for competitive antagonists of the AChR is that they harbor moieties chemically similar to ACh, which select among chemically compatible acceptor groups in the vicinity of the binding site.
The overall results illustrate how a peptide toxin can be used to probe the unknown structure of a neurotransmitter binding site. The findings establish the orientation of CTx ImI as it bridges the ␣ 7 binding site interface, with the N-terminal loop oriented toward the (ϩ) face of the binding site and the Cterminal loop oriented toward the (Ϫ) face. The results also place four of the seven known binding site loops of ␣ 7 in close proximity to bound CTx ImI. The overall theme for ligand binding is a primary anchor through an aromatic-quaternary interaction, which is supplemented by multiple weak interactions. The findings represent a starting point for refining our picture of the molecular structure of the ␣ 7 binding site.